Metal‐Support Interactions Reshape Nanoparticle Catalyst Surfaces

Abstract Tuning metal nanoparticle (NP)‐support interactions is a fundamental method to control heterogeneous catalytic reactions. Experimental techniques provide limited insights into how interactions with the support change the structure of NPs. In contrast, density functional theory (DFT) calculations suffer from system size constraints and rely on idealized constructions. Here we show that the simulation of supported NPs is possible in quantitative agreement with benchmark microcalorimetric measurements for adhesion, chemical potential, and heat of adsorption. Coordination number (CN), strain, and crystallinity analyses in optimized Ag NPs supported on graphene/Ni(111) reveal surface features that deviate from idealized models due to the restructuring of corners and edges. Our results demonstrate that the widely used Wulff construction is not a satisfactory structural model of supported Ag NPs smaller than 8 nm. We suggest replacing Wulff construction with the Winterbottom construction in all future simulation work using idealized models, which reproduce experimental results best, although they still has disagreement with measurements below 5 nm. The direct optimization approach developed here, based on machine learning interatomic potentials (MLIPs), opens up an avenue to simulate industrially relevant supported NPs directly under catalytic conditions without limitations of idealized models.